The task of machine condition monitoring is to detect machine failures at an early stage such that maintenance can be carried out in a timely manner. In the case of failure, it is very important to know the cause of this failure so that corresponding localized, and thus more efficient, maintenance can be applied.
Rule-based systems are perhaps the most widely used condition monitoring approaches. The general format of a rule is “if a condition, then a fault type.” Rules are defined by experts who possess the knowledge of the underlying system model; however, designing accurate rules is a very deliberate and time consuming process, especially for complex systems with many sensors and fault types. For example, it required 80 man years to develop one commercially successful condition monitoring rule base.